Ultrasound-assisted extraction of pectin from artichoke by-products. An artificial neural network approach to pectin characterisation

Artichoke (Cynara scolymus L.) by-products can be used as a good source of pectin. The aim of this work was to compare different pectin extraction methods: power ultrasound (US), enzymes, combination of US and enzymes (US + E), and acids (nitric and sodium citrate). After 6 h, pectin yield was higher when US was applied in combination with Celluclast®1.5 L (up to 13.9%). Structural characterisation showed that US-extracted pectins had lower weight-average molecular weight (Mw) values (146–155 kDa) than pectin extracted with US + E (160–267 kDa) and acid-extracted pectin (329–352 kDa). Monomeric composition reflected that pectin extracted with acids had the highest galacturonic acid (GalA) contents (82.2–90.2%) and the lowest degree of branching [Rha/GalA] (0.026–0.031). Structural characteristics of the different pectins were modelled using two artificial neural networks (ANN) considering composition parameters (Model I) and pectin FT-IR spectra (Model II). In addition, a third ANN was built to determine differences and similarities in the GC-MS spectra of monomeric composition (identified and unidentified monosaccharides) (Model III) showing characteristic patterns with high accuracy rates (above 95% on the test set). Structural differences depending on the extraction method of pectin have been established and these models could be applied to pectin from other sources.This work has been funded by MICINN of Spain, Projects AGL2014-53445-R and AGL2017-84614-C2-1-R. Carlos Sabater thanks his FPU Predoc contract from Spanish MECD (FPU14/03619).Peer reviewe